■Konno, T., Konno, K., Fujimoto, T., Chiba, N.,
Feature Line Extraction and Matching for Modeling Artificial Buildings
Using Measured Point Clouds,
Society for Art and Science, Vol.5, No.3, pp.80-91, (2006)．
As one of modeling methods of representing the real world into virtual space, registration methods that automatically
match a set of measured point clouds have been proposed.
Point cloud registration methods are classified into point-based methods and feature extraction methods.
Although the point-based methods are more accurate than the feature extraction methods, the calculation cost of the registration is large.
The feature extraction methods have a lower calculation cost, but the extraction of the coincident features is very difficult.
In this paper, we propose a registration method that extracts feature lines and matches point clouds measured from different directions.
Because feature lines are extracted, this method is suitable to model buildings, since they naturally unveil linear features.
After the extraction, our method efficiently matches a set of measured point clouds by applying the geometrical transformation, so that the feature lines coincide.
We demonstrate that our method works well for point clouds obtained by measuring real buildings using a range sensor.